Cost-effectiveness of golimumab for the treatment of patients with moderate-to-severe ulcerative colitis in Quebec using a patient level state transition microsimulation

Abstract

Objective: To conduct cost-effectiveness analyses comparing the addition of golimumab to the standard of care (SoC) for treatment of patients with moderate-to-severe ulcerative colitis (UC) who are refractory to conventional therapies in Quebec (Canada).

Methods: An individual patient state transition microsimulation model was developed to project health outcomes and costs over 10 years, using a payer perspective. The incremental benefit estimates for golimumab were driven by induction response and risk of a flare. Flare risks post-induction were derived for golimumab from the PURSUIT maintenance trial and extension study, while those for SoC were derived from the placebo arms of the Active Ulcerative Colitis Trials (ACT) 1 and 2. Other inputs were derived from multiple sources, including retrospective claims analyses and literature. Costs are reported in 2014 Canadian dollars. A 5% annual discount rate was applied to costs and quality-adjusted life-years (QALYs).

Results: Compared with SoC, golimumab was projected to increase the time spent in mild disease or remission states, decrease flare rates, and increase QALYs. These gains were achieved with higher direct medical costs. The incremental cost-effectiveness ratio for golimumab vs SoC was $63,487 per QALY.

Limitations: The long-term flare projections for SoC were based on the data available from the ACT 1 and 2 placebo arms, as data were not available from the PURSUIT maintenance or extension trial. Additionally, the study was limited to only SoC and golimumab, due to the availability of individual patient data to analyze.

Conclusion: This economic analysis concluded that treatment with golimumab is likely more cost-effective vs SoC when considering cost-effectiveness acceptability thresholds from $50,000–$100,000 per QALY.

Keywords:

• Health economics
• ulcerative colitis
• golimumab
• cost-effectiveness
• state transition model

Introduction

Ulcerative colitis (UC) is an inflammatory bowel disease characterized by chronic inflammation of the colon. The clinical course is unpredictable; marked by alternating periods of exacerbation and remission; and can significantly interfere with patients’ daily functioning and quality-of-life^1,2. The prevalence and incidence of UC in Canada rank among the highest in the world. There are ∼104,000 Canadians (0.3% of the total Canadian population) currently living with UC; with an annual incidence rate of 12.9 per 100,000 people. In Quebec, specifically, there were estimated to be ∼24,000 UC cases in 2012³.

Current guidelines recommend tumor necrosis factor (TNF) inhibitors for the treatment of moderate-to-severe UC⁴. Currently, infliximab, golimumab, vedolizumab, and adalimumab have regulatory approval for the treatment of UC from Health Canada. Randomized controlled trials have shown that these biologics induce and maintain response and remission among patients with moderate-to-severe UC who have become refractory to, or who are intolerant of, the standard of care (SoC). We felt it was most appropriate to use the specific standard of care as defined in the clinical trial NCT00385736 (which represented the formulary in Quebec); i.e. a combination of conventional therapies (5-aminosalicytes, corticosteroids, and immunosuppressants)^5–9. All of Canada’s public formularies require a patient to fail treatment on a combination of these conventional therapies prior to reimbursement of biologics; thereby, equating conventional therapies as the appropriate standard of care in economic analyses. However, access to biologics through the public insurance programs varies in each province. Golimumab does have public coverage in the majority of Canada’s provinces. Traditionally, in Quebec, branded biologics (including golimumab) have been reimbursed only as exceptions.

Recently, the Government of Quebec joined the rest of Canada in negotiating non-transparent drug prices; this has resulted in the listing of biosimilar infliximab and vedolizumab on their formulary. However, funding for branded infliximab, adalimumab, or golimumab remains on an exception basis only. In order to access these non-funded biologics, a patient must first fail treatment on biosimilar infliximab and vedolizumab. This requirement to fail treatment on the funded biosimilars may not be appropriate for all patients, since these drugs are intravenously infused. Some patients, especially those for whom commuting to infusion centers is burdensome, could benefit from subcutaneous delivery options. Moreover; the exclusion of biologics on formulary appears to be primarily driven by costs. This lack of funding for biologic golimumab and the sparseness of existing literature together warrant the publication of a cost-effectiveness analysis to aid decision-makers in future formulary decisions^10–13.

Many patients struggle with the disease; cycling and re-cycling through various treatments in an attempt to avoid surgery and complications of the disease. Since most economic models for UC do not capture this treatment cycling^14–16; there is a significant need for a health economic analysis in UC that could consider more realistic treatment patterns. This would permit the real-world accumulation of resource utilization and the costs incurred in the use of anti-TNF agents.

The objective of this study was to assess the cost-effectiveness of golimumab compared with standard therapy within a Canadian setting and payer perspective for patients with moderate-to-severe UC who are non-responsive to conventional therapy. This is the first model based on PURSUIT developed by conducting specific analyses utilizing individual patient data of both of the PURSUIT maintenance trial and also the long-term follow up in the extension study^8,17. It is also the first published microsimulation for UC that considers health states where disease flares are managed prior to patients’ undergoing surgery.

Methods

Model overview

An individual patient state transition microsimulation model was developed to examine the economic (direct and indirect) and clinical outcomes over a 10-year time horizon when comparing golimumab to SoC for a hypothetical cohort of patients with moderate-to-severe UC who had an inadequate response to conventional therapy (Figures 1(A) and (B)). The health states include remission, mild disease, moderate-to-severe disease, post-colectomy, and death. The model structure is broken into two parts: the induction decision tree and the post-induction state transition model.

Figure 1. Patient flow for induction decision tree and post-induction state transition model. (A) Induction decision tree. (B) Post-induction state transition model.

Induction decision tree

As the simulation begins; the cloned patients enter the model in the moderate-to-severe disease state and being treated with SoC. SoC is defined as a combination of 5-aminosalicytes, corticosteroids, and immunosuppressants used by subjects in the PURSUIT trial. The conventional therapies themselves are not modeled individually, but rather the distribution of SoC treatments remains constant for both arms over the model time horizon. (Additional details are provided in the Comparator efficacy and adverse events section and Table 1). Next, they move to an induction period during which the patient receives the appropriate dosage of the assigned comparator (either golimumab plus SoC, or SoC alone). The probability of a response and remission are conditional on the treatment received during induction. At the end of the induction period, patients either respond to the treatment and transition to the remission or mild disease states, or do not respond and remain in the moderate-to-severe disease state.

Table 1. Cost inputs.

	Parameter	Cost	Source
Golimumab	Induction 200 mg Week 0 & 100 mg Week 2	$4,341.00 per induction	RAMQ^38
	Maintenance 100 mg every 4 weeks	$1,447.00 per administration	RAMQ^38
SoC	Weighted average of 6-MP, 5-ASAs, MTX, and steroids	$1,248.69 per year^a	RAMQ^38
Flare Management	Outpatient management	$534.00 per flare^b	RAMQ^39^,^40
	Inpatient stay (No colectomy)	$10,169.00 per stay^c	supplemental material
Colectomy-related	Colectomy	$42,422.00 per patient^c	supplemental material
	Acute pouchitis	$10.51 per year^d	RAMQ^38
	Chronic pouchitis	$3,841.10 per year^e	RAMQ^38
Adverse event	Minor infection	$68.19 per event^f	RAMQ^38^,^40; CDM^41
	Serious infection	$5,345.96 per event^g	CIHI^42

5-ASA, 5-aminosalicylic acid; MTX, methotrexate; SoC, standard of care.

^a Weighted cost was calculated based on the fact that 31.20% of patients received 6-mercaptopurine (1.5 mg/kg × average weight of PURSUIT trial patients [73.54 kg]), 81.90% received 5-aminosalicylic acid (6 mg per day), 1.20% received methotrexate (25 mg per day), and 42.80% received steroids (50 mg per day).

^b Assumption that includes two gastrointestinal specialist visits, one anti-nuclear antibody test, one anti-deoxyribonucleic acid antibody test, and one colonoscopy without polypectomy.

^c Mean inpatient stay costs accrued per patient from a retrospective analysis of the hospital admissions in the Med-Echo database from an UC cohort in Quebec, see the supplemental material for details.

^d 5-aminosalicylic acid cost at a dosage of 6 mg per day.

^e Mercaptopurine cost at a dosage of 1.5 mg/kg/day using the average weight of PURSUIT trial patients (73.54 kg).

^f Includes one general practitioner visit and oral clarithromycin at a dosage of 500 mg twice a day for 7 days.

^g Weighted average of inpatient stay cost for bacterial pneumonia and viral/unspecified pneumonia (CMG codes: 136 and 138).

Post-induction state transition model

The state transition model has four health states (remission, mild disease, moderate-to-severe disease, and post-colectomy) plus an absorbing death state. For each 4-week cycle, patients can remain in the same health state, transition to another health state, or transition to the death state; however, a patient who transitions into the post-colectomy state remains in this state for the rest of their time alive in the simulation. Transitions from remission, mild disease, or moderate-to-severe disease to other health states are triggered by two possible events: (a) a flare; or (b) a non-emergent colectomy. A flare is an indicator of increased disease severity that typically results in the reappearance of symptoms of moderate-to-severe disease. The probability of a flare is conditional on both the health state and treatment. Flares result in a change in the patient’s management and potentially change in treatment, as described in the Flare management section. Following the management of a flare that does not result in a colectomy; a patient is assigned back to one of three health states (remission, mild, or moderate-to-severe) depending on their response during flare management, and the patient resumes SoC. This cycling continues until a patient has a colectomy, whether flare-related or non-emergent, or dies, or the end of the model time horizon is reached. The probability of death is conditional on age, but not health state. Patients are also at risk of other events that do not lead to a change in health state: adverse events (injection site reactions and minor or serious infections) or discontinuation of maintenance therapy (the risk of a flare is then updated).

Post-induction maintenance treatment

Patients receiving golimumab plus SOC during the induction period who respond will continue to receive that therapy during maintenance; while non-responders stop the anti-TNF and revert to only SoC. Patients who were on SoC alone continue on SoC, regardless of the response at the end of the induction period.

Flare management

Upon beginning flare events, patients who were receiving the golimumab plus SoC treatment must stop their golimumab. A flare can be managed in either an inpatient or outpatient setting. Patients who are treated in the outpatient setting with SOC either respond to treatment and transition to the remission or mild disease state, or do not respond and transition to the moderate-to-severe disease state. Among those who do not respond to treatment, some are classified as in a flare, are moved to an inpatient setting where they receive additional management, and are subsequently discharged on SoC.

All other patients receive inpatient care and enter a decision tree based on the 2012 Toronto UC Guidelines; containing multiple decision nodes and assessments of response¹⁸. Ultimately, flare patients who were receiving golimumab plus SoC treatment either (a) exit the decision tree in the mild state, stopping golimumab and continuing maintenance with SoC only; or (b) receive a colectomy and transition to the post-colectomy state. Following year 4 within the model, patients are no longer at risk of receiving a colectomy.

Colectomy

Patients transition to the post-colectomy state following either (a) the management of a flare that led to colectomy, or (b) a non-emergent colectomy event. A non-emergent colectomy is defined as a colectomy that does not occur due to a flare (i.e. patients may elect to have a colectomy). Based on clinical input, it is assumed that patients are at risk of a non-emergent colectomy if they are (a) non-responders in the moderate-to-severe state after induction, or (b) have a history of a flare (i.e. patients with a post-induction loss of response).

Utilities

The model utilities are assigned by health state (i.e. remission, mild disease state, moderate-to-severe disease state, post-colectomy without pouchitis) and also for specific events (i.e. flare or chronic pouchitis), and are accumulated over each cycle. Adverse events are assumed to be transient, and do not impact quality-of-life in this model.

Costs

The direct medical costs accrued in the model include costs associated with drug acquisition, management of flares, colectomies, complications of colectomies, and adverse events (see Table 1 for details). Drug acquisition costs incorporate only those drugs administered outside of the inpatient setting (induction, maintenance, or outpatient flare management), as those administered in the inpatient setting are captured in the cost of that event. Flares were managed in two different settings: either in an outpatient facility or an inpatient hospital setting. Flares managed in an outpatient setting are assumed to be of a milder variety and accrue a cost for the outpatient facility fee. Flares that resulted in an inpatient admission accrue the cost of an inpatient stay for either a general hospitalization or colectomy surgery, depending on their final management decision. Complications due to colectomy were both short-term (acute pouchitis) and long-term (chronic pouchitis); with costs for acute pouchitis accruing only at the time of the event; while costs for medical management of chronic pouchitis accrue for the remainder of the time in the model.

The societal perspective scenario analysis examined three components of indirect costs: lost productivity, short-term work loss due to an event, and long-term work loss. Lost productivity due to reduced working capacity for employed patients is dependent on the health state. In addition, short-term work losses (such as sick days) due to a colectomy or inpatient flare management are accrued. Long-term work loss is estimated only for patients in the post-colectomy state. Additionally, changes in employment status are captured only for patients in the post-colectomy state, where it is assumed that patients who have undergone a colectomy have a risk of losing their employment for the first 3 years post-colectomy. Additional details are provided in the supplemental material.

Outputs

Model outputs include quality-adjusted life-years (QALYs), direct, total costs, and incremental cost-effectiveness ratios (ICERs) calculated using incremental total costs per incremental QALY gained. Additional clinical outcomes are reported, including response and remission achievement, time in response or remission, and event counts (flares, hospitalizations, and colectomies). Additional outputs are included for the societal perspective scenario, including indirect costs, total (direct plus indirect) costs, and the societal perspective cost per incremental QALY gained.

Data sources

Model inputs and data sources are summarized in Tables 1 and 2. All health utilities in the model were derived from a publication by Woehl et al.¹⁹; except for pouchitis²⁰. All costs in the model are reported in 2014 Canadian dollars. Direct medical cost inputs and sources are shown in Table 1. Drug costs were based on the wholesale acquisition cost per package obtained from the List of Medications published by RAMQ (Régie de l’assurance maladie du Québec); using the generic drug acquisition cost whenever available.

Table 2. Model Inputs

Parameter		Treatment	Input Value	Source
Induction	% Responders	Golimumab	60.07%	^6^,^21
		SoC	37.30%
	% Remission	Golimumab	36.55%
		SoC	14.90%
Adverse Events per Cycle	Injection Site Reaction	Golimumab	0.80%	^6–8
		SoC	1.55%
	Minor Infection	Golimumab	1.07%
		SoC	2.53%
	Serious Infection	Golimumab	0.00%
		SoC	0.46%
Post-Induction Flare	See supplemental material; Table S2; for details of Weibull functions			^6^,^8^,^17
Cohort Baseline Age; Mean Years (SD)			40.00 (13.37)	^7
Location of Flare Management; % Inpatient Management			22.91%	Unpublished^a
% of Moderate-to-Severe Disease State in a Flare End of Induction			29.31%	^8
% Responders: Inpatient First-Line Treatment			75.00%	Assumption^b
% Responders: Inpatient Second-Line Treatment			84.21%	^38
% of Inpatients with Immediate Colectomy			33.52%	Unpublished^a
Monthly Non-Emergent Colectomy			0.51%	Unpublished^a
Post-colectomy: Acute Pouchitis			32.00%	^39
Post-colectomy: Chronic Pouchitis			25.00%	Assumption
Health Utilities
Remission			0.87	^19^,^20
Mild Disease			0.76
Moderate-to-Severe Disease or Flare			0.41
Post-colectomy: Without Pouchitis			0.70
Post-colectomy: With Pouchitis			0.43

Abbreviations. SD, Standard deviation; SoC, standard of care.

^a Retrospective database analysis performed by Evidera, using the MarketScan Commercial Claims and Encounters, and the Medicare Supplemental and Coordination of Benefits databases, 2008–2011, from Truven Health Analytics. Treatment Patterns and Resource Use for Patients with Ulcerative Colitis Who Have Failed Treatment with Infliximab. Janssen internal data; not published. Refer to the supplemental material (Flare Management and Colectomy) for details.

^b Assumption. Refer to the supplemental material (Flare Management and Colectomy) for details.

Inpatient costs were based on a retrospective analysis of the hospital admissions in the Med-Echo database from an UC cohort in Quebec between 2005–2011; additional details are provided in the supplemental material.

Indirect cost input sources (productivity losses; short- and long-term work losses) used in the societal scenario analysis are detailed in the supplemental material; Table S4.

Comparator efficacy and adverse events

A published network meta-analysis²¹, based on data from randomized, placebo-controlled trials that included golimumab (PURSUIT), was used as the source for the relative efficacy of the comparators at induction vs SoC. SoC induction efficacy was from the ACT 1 and ACT 2 trials⁶. Efficacy for SoC was taken from alternative sources because of the two-part design of PURSUIT. The first part of PURSUIT, the PURSUIT Induction⁷, compared golimumab with placebo for the induction of response and remission at Week 6. Only those patients who received golimumab and who at least responded to induction moved on to the second part of PURSUIT, PURSUIT Maintenance⁸. Patients were randomized to either golimumab or placebo for the 54-week maintenance study. Because of this design, there is no contiguous placebo (SoC) arm to follow for clinical data in PURSUIT. Each of these trials enrolled patients with moderate-to-severe active ulcerative colitis undergoing treatment with conventional medications (5-aminosalicylates, corticosteroids, or azathioprine, 6-mercaptopurine, and/or methotrexate). The probability of remission or mild disease after induction is shown in Table 2. The Mayo Clinic score, a scoring system for assessment of UC activity, was used to define the remission or mild disease state. Remission is defined as a Mayo score ≤2 points with no individual sub-score >1. Patients with a clinical response who did not meet the criteria for remission are considered to be in the mild disease state. Clinical response is defined as the decrease from baseline Mayo score by ≥30% and ≥3 points; with either a decrease from baseline in the rectal bleeding sub-score of ≥1 or a rectal bleeding sub-score of 0 or 1.

The flare risk for the first year post-induction was based on an analysis of the PURSUIT Maintenance trial⁸. The projections for subsequent years for golimumab were developed by fitting the data available from the PURSUIT extension trial¹⁷. The long-term projections for SoC were based on analysis of the first year post-induction using the data available from the ACT 1 and 2 placebo arms, as the PURSUIT extension trial did not include a placebo arm. The flare probabilities for the first-year post-induction and later years were derived from Weibull survival functions fitted to these trials. The flare definitions differed among the PURSUIT trial and its extension study, and more details of the derivation are provided in the supplemental material. Within the first year post-induction, a flare is defined as a two-point increase in the partial Mayo score or a total partial Mayo score of ≥7, while the extension study used the Physician Global Assessment (PGA) to define a flare as an increase of at least two PGA points in subsequent visits or initiation of corticosteroids, a UC-related hospitalization or surgery, or discontinuation of study drug due to lack of efficacy during the extension study.

Adverse event probabilities for SoC were based on the ACT 1 trial⁶ and for golimumab the PURSUIT trial (Table 2)⁸.

Flare management and colectomy

The probabilities for an inpatient admission vs outpatient management of a flare and undergoing a colectomy were based on an analysis of US claims data from 2008–2011 (MarketScan Commercial Claims and Encounters, and Medicare Supplemental and Coordination of Benefits databases, available from Truven Health Analytics, 2008–2011). To derive these estimates, the analysis assumed that stopping treatment with infliximab, another TNF inhibitor marketed for the management of UC, was a reasonable proxy for flare. The inpatient decision tree was calibrated to match the probability of a colectomy estimated from the US claims data, this was because Canadian-specific data for this population was not available for this input. The probability of a non-emergent colectomy was also derived from the same claims data analysis and was defined as a colectomy occurring more than 1 year after the initial infliximab discontinuation. Refer to the supplemental material for more details.

Mortality

Age-specific mortality data for the Canadian general population²² were used to derive a Gompertz distribution (supplemental material; Table S3). It is assumed that neither UC nor a colectomy increases the risk of mortality.

Analyses

The base-case evaluated moderate-to-severe UC patients who are refractory to conventional therapies assuming golimumab 200 mg at Week 0 and golimumab 100 mg at Week 2 (induction), followed by golimumab 100 mg every 4 weeks (maintenance) combined with SoC, versus SoC alone. A scenario analysis is also presented with an alternative dose regimen (50 mg maintenance dose, the cost of golimumab 50 mg is the same as 100 mg). These dosing schedules were based on the recommended doses per the Health Canada Product Monograph.

The base-case used the payer perspective for Quebec (Canada), considering only direct medical costs over a 10-year time horizon, as the efficacy is extrapolated from the PURSUIT maintenance and extension trial, and this time horizon is long enough to accumulate all cost and health differences across comparators. Costs and clinical outcomes were discounted at an annual rate of 5.0%, in accordance with Canadian guidelines for economic evaluations²³. A scenario analysis was conducted to understand the results from a societal perspective, as this is the perspective used by the health authority within Quebec when making formulary decisions.

A series of one-way sensitivity analyses and probabilistic sensitivity analyses were conducted. (For details on the probabilistic sensitivity analyses, see Table S5 in the supplemental material.) For these analyses, maintenance therapy was assumed to be stopped when patients experienced a flare. A series of scenario analyses are presented (Table 3) that include an alternative golimumab maintenance dose regimen, time horizons, and utility values for all health states using the values reported in Swinburn et al.²⁴, a study of 230 patients with UC (30 of whom were post-surgery) assessing their EQ-5D and their disease severity. The flare rates associated with a 50 mg maintenance dose have also been derived from an analysis based of the PURSUIT maintenance and extension trial data.

Table 3. Scenario analyses results: golimumab vs SoC.

Scenario	Net QALYs	Net direct costs	Incremental cost/QALY	Change (%) vs base-case
Base-case	0.55	$35,363	$63,847	—
Societal perspective	0.55	$35,363 + ($9,177); indirect	$47,277	−25.95%
Golimumab 50 mg maintenance dose	0.34	$26,404	$78,458	22.88%
Time Horizon: 1 year	0.11	$10,614	$96,417	51.01%
Time Horizon: 5 years	0.39	$24,909	$63,874	0.04%
Time Horizon: 20 years	0.64	$57,464	$89,438	40.08%
Swinburn Utility*	0.53	$35,363	$66,722	4.50%

Abbreviations. QALY, quality-adjusted life-year; SoC, Standard of care.

* Moderate-to-severe or flare 0.55; mild 0.80; remission 0.91; and post-colectomy 0.59.

Results are based on 20 replications of 1,000 patients per cohort.

Results

Base-case analysis

Base-case results are presented in Table 4 and show clinical benefits and costs from a payer perspective. Over a 10-year time horizon, SoC resulted in 7.97 life-years, 3.99 QALYs, with total direct costs of $42,158. Patients receiving SoC spent, on average, 3.14 years in the mild or remission health states and just 0.80 years in remission. Golimumab increased the time spent in the mild disease or remission states (1.48 years), and also lowered the flare counts and associated colectomies, this improvement in health outcomes translated into an increase to 4.54 QALYs. Over 10 years, on average, patients receiving SoC experienced 4.91 flares, leading to 1.70 hospitalizations, while those on golimumab had many fewer flares (−0.80), and associated hospitalizations (−0.31). While many patients on SoC had a colectomy (51%), this risk was also somewhat reduced to 40% by golimumab.

Table 4. Base-case analysis results.

Results per patient	Golimumab	SoC	Net difference Golimumab vs SoC
Clinical outcomes
Mean LYs (Discounted)	7.97	7.97	0.00
Mean QALYs (Discounted)	4.54	3.99	0.55
Mild disease or remission states (Undiscounted)
% at Induction	60%	37%	22%
% at Week 54	37%	17%	20%
Remission state (Undiscounted)
% at Induction	37%	15%	22%
% at Week 54	25%	8%	17%
Time in disease state (Years; Undiscounted)
Mild disease or remission	4.63	3.14	1.48
Remission	2.15	0.80	1.35
Event counts (Undiscounted)
Flares	4.11	4.91	−0.80
Hospitalizations	1.39	1.70	−0.31
Colectomies	0.40	0.51	−0.11
Economic outcomes
Direct costs (Discounted)	$77,520	$42,158	$35,363
Active treatment	$41,949	$0	$41,949
SoC	$9,867	$9,867	$0
Flare	$11,495	$14,102	−$2,607
Colectomy	$14,205	$18,102	−$3,897
Adverse events	$5	$87	−$83
Incremental results
Incremental direct costs per QALY (ICER)
Public payer perspective			$63,847

Abbreviations. ICER, incremental cost-effectiveness ratio; LY, life-year; QALY, quality-adjusted life-year; SoC, standard of care.

Results are based on 20 replications of 1,000 patients per cohort.

Total direct costs were higher for patients treated with golimumab, with a total of $77;520. When compared to SoC; golimumab was associated with incremental QALY gains and additional direct medical costs, leading to an ICER of $63,847 from the payer perspective.

Scenario sensitivity analyses

Total indirect costs for SoC were $78,785, while they were lower for golimumab ($49,608), leading to ICERs of $47,277 per QALY. Treatment of patients with golimumab at a lower maintenance dosage (50 mg) resulted in a moderate increase (∼ 23%) in the ICER due to the lower QALYs per patient (0.34 QALY with golimumab 50 mg vs 0.55 QALY with golimumab 100 mg). Conducting analyses with the alternative utility values from Swinburn et al.²⁴ had little impact on the overall results. While the moderate-to-severe utility value from Swinburn et al.²⁴ was higher than that used in the base case, the lower post-colectomy utility offset the resulting decrease in the incremental QALYs when compared to SoC.

Deterministic and probabilistic sensitivity analyses

Figure 2 presents the results from the deterministic sensitivity analysis on the base-case for the top 14 parameters for golimumab that had the most impact on the ICERs. These analyses found the results were most sensitive to varying utility values for remission, post-colectomy, and moderate-to-severe disease. As would be expected, the maintenance costs and response induction were also important drivers for the results.

Figure 2. Tornado diagram for ICERs: Golimumab vs SoC. The base-case ICER for the deterministic sensitivity analysis for golimumab ($69,966/QALY) is slightly different from the ICER reported for golimumab in the Results section for the base-case analysis ($63,487/QALY), as a lower number of patients and replications were used in the deterministic sensitivity analysis (five replications of 500 patients per treatment cohort) compared to what was used in the real base-case scenario for golimumab (20 replications of 1,000 patients per treatment cohort). One-way sensitivity analyses were conducted by increasing or decreasing base-case values by 20%, except for the annual discount rates for costs and health outcomes, which were set to 0% and 3%, respectively. Results of the analyses evaluating lower and upper bounds for health and cost discount rates are on the same side of the base-case ICER, as the lower bound was set to 0% and higher bound to 3%. GOL, golimumab; ICER, incremental cost-effectiveness ratio; QALY, quality-adjusted life-year; SoC, standard of care.

Probabilistic sensitivity analyses on each base-case were also conducted; the cost-effectiveness acceptability curve for golimumab is presented in Figure 3. The model tested the willingness-to-pay threshold at which golimumab would be a more cost-effective strategy than SoC. Golimumab was a more cost-effective strategy than SoC at a willingness-to-pay threshold of $63,000 per QALY or greater. At $50,000 per QALY, golimumab was considered to be a cost-effective treatment in 29% of the iterations. At a higher willingness-to-pay threshold of $100,000/QALY; 75% of the iterations showed golimumab was a cost-effective alternative to SoC. Within a cost-effectiveness threshold from $50,000–$100,000, the results of the probabilistic sensitivity analysis indicate with moderate certainty that golimumab is likely a cost-effective treatment.

Figure 3. Cost-effectiveness acceptability crve: Golimumab vs SoC. Probabilistic sensitivity analyses were conducted in which all key parameters were simultaneously varied according to the assigned probability distribution. When available, standard errors from the parameter source data were used to measure parameter uncertainty, otherwise 20% of the mean value was assumed to be the standard error.

Discussion

This economic evaluation illustrates that golimumab induction along with maintenance therapy may improve health outcomes, as more patients were projected to continue to have mild disease or remain in remission. Since there is assumed to be no difference in survival between treatment arms, differences in QALYs and other clinical outputs are driven by time in different disease states.

As a result of this improvement, patients had a lower risk of experiencing a flare and its associated hospitalizations and colectomies, as has been demonstrated in the real-world data analysis conducted by Van der Valk et al.²⁵. Based on our analyses, over a decade, patients with moderate-to-severe UC managed with SoC can, on average, be expected to experience about five flares, and half of those patients will undergo a colectomy. In the base-case analyses, one-third of the direct medical costs estimated for patients on SoC over a decade were associated with inpatient or outpatient flare management, and these were reduced by ∼17% with golimumab. The available evidence in the literature indicated that work productivity of patients with UC was also highly affected^3,26–32. In our scenario analysis examining the societal perspective, the indirect costs were substantial, exceeding the direct medical costs by almost 2-fold for the cohort receiving SoC. The indirect costs, therefore, accounted for the majority of the total costs (65%) with SoC, and were estimated to be reduced by 11% with golimumab. We do add a caveat that there is some uncertainty in the estimates.

The published cost-effectiveness analyses for anti-TNFs varies widely in the literature, ranging from $64,000 per QALY in Ung et al.¹⁴ to $358,000 per QALY in Xie et al.¹⁵; both of these analyses were conducted in Canada. The main difference between the aforementioned analyses was the treatment algorithm: Xie et al.¹⁵ allowed patients to cycle into surgery upon loss of response to an initial anti-TNF or SoC; whereas Ung et al.¹⁴ incorporated real-world clinical experience to reflect patient preferences to remain in a chronically unwell state and to delay surgery. The simplification of the treatment algorithm by Xie et al.¹⁵ likely explains the higher ICERs for anti-TNFs, since patients quickly cycled into surgery (and post-surgery), where their accumulated costs were relatively low compared to the cost of biologics, and their quality-of-life was assumed to be as good as in remission on therapy. This analysis took a novel approach of creating a state transition microsimulation model where patients were able to manage their flares, which we intended to more closely reflect actual clinical practice.

While several network meta-analyses reported that infliximab is believed to be the most efficacious anti-TNF in UC, we would argue that the incremental QALYs gained with golimumab are comparable to those of infliximab. When comparing QALYs to a study by Tsai et al.³³, referenced in the NICE review of biologics in UC³⁴, it was determined that, over 10 years, infliximab resulted in incremental QALYs vs standard of care of 0.387–0.753 for remitters and responders, respectively. The incremental QALY gained for golimumab over 10 years is 0.55, clearly in line with that of infliximab.

Recently, the biosimilar infliximab has been added to formularies with a significant price discount vs branded infliximab; this is likely to greatly improve the cost-effectiveness of infliximab in UC. Biosimilar infliximab is likely to be a more cost-effective treatment option relative to golimumab, given its significant price discount and assuming comparable efficacy to branded infliximab. Regardless, there is a need for subcutaneous options that can be taken from home for UC patients who cannot access infusion clinics.

Previously, economic analyses in UC compared surgery versus treatment with biologics. Archer et al.¹⁰ compared elective surgery with anti-TNF treatment, and these economic analyses indicated that colectomy could dominate drug therapies. However, in the latest guidance from NICE in the UK for anti-TNF treatment in UC³⁴, it was stated that, following the opinion of clinicians, elective colectomy is not an appropriate comparator, because, even if surgery may be curative, it does not fully restore the patient’s quality-of-life¹⁰.

This opinion was emphasized, especially as it applied to young people, given that the surgery could adversely affect a patient’s ability to fully participate in life events, e.g. be unable to have children³⁴. The professional groups further noted the difficulty of comparing the quality-of-life of patients treated with TNF inhibitors such as golimumab vs those who have colectomy, given the strict inclusion criteria of clinical trials³⁴.

As with all models, some assumptions and limitations were necessary to support long-term projections. The decision to include patient-level analyses of PURSUIT and ACT 1 and 2 trial data constrained our ability to include other treatments available in the treatment of moderate-to-severe UC, which is a limitation of the study, given the changing treatment landscape. Projections of flares in both short-term for non-responders and long-term flare projections for SoC were based on the data available from the ACT 1 and 2 placebo arms, as the relevant data were not available from the PURSUIT maintenance or extension trial, respectively. While this introduces bias into the projections, the PURSUIT and ACT 1 and 2 trials had many of the same inclusion and exclusion criteria^6,8, as well as similar baseline characteristics and demographics^35,36. Unlike other published models, the flare rates on golimumab maintenance therapy were derived from the long-term follow-up collected by analyzing both the PURSUIT maintenance and extension studies. Analyzing these long-term follow-up studies may have allowed this model to more appropriately capture the impact of maintenance therapy than other models. However, deriving flare rates from the ACT 1 and ACT 2 placebo arms, and the extrapolation of a single arm extension trial data for long-term flare rates, may also have introduced a bias. The generalizability of these long-term, trial-based estimates for flare risks is also uncertain because they were unavailable from real-world databases in Quebec. A limitation of this model is that there is no comparison to other TNF inhibitors or biosimilars approved for UC. However, given that reimbursement of these biologics is limited, it is appropriate to compare to conventional medications as the current standard of care. We acknowledge that, strictly speaking, the generalizability of this analysis is limited to Quebec, given that we used Quebec-specific inputs where possible (e.g. costs, resource use). However, this analysis may have some relevance to other regions that allow the reimbursement of biologics only as exceptions, this may include countries in Europe and the Middle East. For these analyses, patients responding at the end of the induction period were assumed to stop maintenance therapy only when they experienced a flare. However, there is evidence that patients remain on active treatment longer than our base-case analysis projected. For example, in routine clinical practice, they may continue with maintenance therapy and add supplementary SoC therapies. The model did not quantify potential health benefits of the steroid tapering/sparing whilst, in clinical practice, patients in remission would taper and aim to discontinue steroids. The cost of SoC was not updated for those with remission, as the cost difference was minimal.

Management of flares contributed the most to the direct medical costs, and the proportion of patients managed in the inpatient setting following a flare, and the frequency of colectomy was based on an analysis of US claims, as no data analysis has been performed for this specific population in either Quebec or Canada. The assumption that stopping infliximab is indicative of a flare in the claims analysis had limitations, because patients could have stopped infliximab due to other reasons such as adverse events.

Our analyses also assumed that there was no risk of an emergency colectomy following year 4. We did this for several reasons. First, we assume that the number of patients undergoing colectomy increases the longer the colectomy risk is present. Clinicians have expressed concern that this compromises model face validity, and have criticized other economic modeling studies for these higher colectomy numbers. Second, the 2008 cohort study by Ferrante et al.³⁷ reported that few patients undergo a colectomy more than 4 years after their diagnosis with UC. Third, the clinical UC experts who worked on this study and manuscript approved the assumption as representative of what is commonly seen in the community. While those clinical experts agree that emergent colectomies do occur over time, they posit that the risk decreases as years of remission increase. However, given the lack of data to appropriately model that pattern, and the complexity it would add to the model, we decided to keep the risk of emergent colectomy flat, especially as this should, in fact, actually bias against golimumab. It was also assumed for these analyses that neither UC nor a colectomy increases the risk of mortality.

Utility inputs for the base-case analyses were obtained from an observational study conducted by Woehl et al.¹⁹, except for the “post-colectomy with complications” input, which was obtained from Arseneau et al.²⁰. Obtaining most of the utilities from a single source provided internal validity to our economic assessment. Furthermore; the utilities from Woehl et al.¹⁹ were more similar to the utility values obtained from an unpublished real-world utility survey of Canadian patients with UC (from Janssen’s BioAdvance® clinics: Janssen internal data only, not published), as opposed to the PURSUIT trial-based utilities. The scenario analyses with the alternative utility values from Swinburn et al.²⁴ illustrated that this had little impact on the overall results.

Conclusions

Our analyses demonstrate that golimumab treatment in patients with moderate-to-severe UC who have had an inadequate response to conventional therapy provides additional clinical benefits for the patients, and still meets the fiscal needs of public payers in Quebec (Canada) when costs of golimumab are compared with those of the current SoC.

Transparency

Declaration of funding

The cost-effectiveness study and this resulting article were funded in full by the manufacturer of golimumab, Janssen Biotech, Inc. Janssen assigned their employees PD and DN to the project, contracted with and paid Evidera for their services on the study and manuscript, and paid honoraria to the seven independent authors for their work on the project.

Declaration of financial/other relationships

AJW, SS, and CSLT are employed by Evidera, Inc. In their salaried positions, they work with a variety of companies and are precluded from receiving payment or honoraria directly from these organizations for services rendered. Evidera received payment from Janssen Biotech of Horsham, PA for the conduct of this study and the development of this manuscript. FC is the Principal at Damos, Inc., and has received payment for serving as a consultant/advisor for Janssen, Inc. Canada. ER is an Associate Professor and researcher at the Research Institute of the McGill University Health Centre. She has received grants and consultant fees from Janssen, Inc. and Bristol-Myers Squibb Canada. GA is a gastroenterologist for CIUSS est de Montreal at the Maisonneuve-Rosemont Hôpital and serves as Professor of Medicine at the Université de Montréal. He is a minor shareholder of 100 shares US of Janssen and has received honoraria/compensation for his work on speaker’s bureaus for Janssen, Takeda, Shire, and Allergan, and as a consultant for Janssen, AbbVie, Takeda, Shire, Allergen, and Lupin. He has received sponsorship from Janssen, Takeda, and Ferring, and clinical research funding from Janssen, AbbVie, Takeda, Ferring, Lilly, Shire, and Genentech. He received honoraria from Janssen for his work on this project. EJB is a gastroenterologist and the Director of the IBD Clinic at the Hôtel-Dieu de Montreal and an Assistant Professor of Medicine at Université de Montréal. He has received grants for the development of educational tools for IBD patients from AbbVie/Shire and for clinical trials from AbbVie. He has received honoraria/compensation for his participation in advisory boards and lectures/presentations from Janssen, AbbVie, Takeda, Shire, Pfizer, and Ferring and received compensation for serving on an advisory board for Merck and a speaker’s bureau for Allergan. He did receive honoraria from Janssen for his work on this project. RB is a gastroenterologist for CISSS Chaudière-Apalaches at Hôtel-Dieu de Lévis and an Associate Clinical Professor at Université Laval. He has received fees for acting as a consultant/advisor and serving on the speaker’s bureau for the following companies: AbbVie, Janssen, Pfizer, Shire, and Takeda. He received honoraria from Janssen for his work on this project. AC is the Director of the Division of Gastroenterology at Jewish General Hospital and serves as an Associate Professor of Medicine at McGill University. He has received honoraria for his work as a consultant/advisor for AbbVie, Janssen, Pfizer, Takeda, Ferring, Allergan, and Pendopharm. PP works for Centre Hospitalier Universitaire de Québec at the Hôpital du Saint-Sacrement in Québec and serves as a Professor of Medicine at Université Laval. He has served as a consultant/advisor for AbbVie Canada, Actavis, and Janssen Canada and on a speaker’s bureau for Janssen Canada. He received honoraria from Janssen for his work on this project. DN is an employee of Janssen Pharmaceutica, NV of Belgium and a shareholder in Johnson & Johnson. PD is an employee and shareholder of Janssen Biotech. Peer reviewers on this manuscript have received an honorarium from JME for their review work, but have no other relevant financial relationships to disclose.

Acknowledgments

The authors would like to acknowledge Sandra Milev from Evidera and Dan Pettitt and Timothy Gathany from Janssen, at the time of study conduct, for their participation in either this research project or the writing of this manuscript. They would further like to thank Janet Dooley of Evidera’s Production Team for her assistance in the preparation and submission of this manuscript. The authors note that there has not been any previous publication of the data in this manuscript.